Native and Domestic Browsers and Grazers Reduce Fuels, Fire

Home , Grazing (behaviour), Tetraponera penzigi

Ecological Applications, 24(4), 2014, pp. 741–749 Ó 2014 by the Ecological Society of America

Native and domestic browsers and grazers reduce fuels, ﬁre temperatures, and acacia ant mortality in an African savanna

1,2 2,3 2,4 2,5 2,6,7 DUNCAN M. KIMUYU, RYAN L. SENSENIG, CORINNA RIGINOS, KARI E. VEBLEN, AND TRUMAN P. YOUNG 1Department of Natural Resources, Karatina University, Karatina 10101 Kenya 2Mpala Research Centre, P.O. Box 555, Nanyuki 10400 Kenya 3Department of Biological Sciences, Goshen College, Goshen, Indiana 46526 USA 4Conservation Research Center, Teton Science Schools, Jackson, Wyoming 83001 USA 5Department of Wildland Resources, Utah State University, Logan, Utah 84322 USA 6Department of Plant Sciences, University of California, Davis, California 95616 USA

Abstract. Despite the importance of fire and herbivory in structuring savanna systems, few replicated experiments have examined the interactive effects of herbivory and fire on plant dynamics. In addition, the effects of fire on associated ant–tree mutualisms have been largely unexplored. We carried out small controlled burns in each of 18 herbivore treatment plots of the Kenya Long-term Exclosure Experiment (KLEE), where experimentally excluding elephants has resulted in 42% greater tree densities. The KLEE design includes six different herbivore treatments that allowed us to examine how different combinations of megaherbivore wildlife, mesoherbivore wildlife, and cattle affect fire temperatures and subsequent loss of ant symbionts from Acacia trees. Before burning, we quantified herbaceous fuel loads and plant community composition. We tagged all trees, measured their height and basal diameter, and identified the resident ant species on each. We recorded weather conditions during the burns and used ceramic tiles painted with fire-sensitive paints to estimate fire temperatures at different heights and in different microsites (under vs. between trees). Across all treatments, fire temperatures were highest at 0–50 cm off the ground and hotter in the grass under trees than in the grassy areas between trees. Plots with more trees burned hotter than plots with fewer trees, perhaps because of greater fine woody debris. Plots grazed by wildlife and by cattle prior to burning had lower herbaceous fuel loads and experienced lower burn temperatures than ungrazed plots. Many trees lost their ant colonies during the burns. Ant survivorship differed by ant species and at the plot level was positively associated with previous herbivory (and lower fire temperatures). Across all treatments, ant colonies on taller trees were more likely to survive, but even some of the tallest trees lost their ant colonies. Our study marks a significant step in understanding the mechanisms that underlie the interactions between fire and herbivory in savanna ecosystems. Key words: Acacia drepanolobium; cattle; Crematogaster; elephants; herbivory; Kenya; Laikipia; livestock; mutualism; Tetraponera.

INTRODUCTION of modeling approaches (van Langevelde 2003, Baxter Fire and herbivory are important drivers in savanna and Getz 2005, Holdo et al. 2009) and empirical studies systems, playing key roles in structuring tree–grass (Dublin et al. 1990, Moncrieff et al. 2008) have coexistence (reviewed in Sankaran et al. 2005, 2008), suggested that fire may interact with grazing and landscape heterogeneity (Fuhlendorf and Engle 2004, browsing in complex ways (Collins and Smith 2006) to Kerby et al. 2007, Waldram et al. 2008, Allred et al. create systems that are dynamic and nonequilibrial 2011), forage quantity and quality (Sensenig et al. 2010), (Bond and Keeley 2005), changing at various scales in and plant productivity (Holdo et al. 2009). Although it space and time (Gillson 2004, Archibald et al. 2005). is well documented that both fire (Mapiye et al. 2008, Understanding the mechanisms underlying how fire Levick et al. 2009, Winston and Trollope 2011) and and herbivory interact with changes in tree cover in herbivory (McNaughton et al. 1988, Levick et al. 2009) particular has both important theoretical and applied independently affect savanna ecosystems, there is implications, since variation in savanna woody cover increasing interest in understanding their potentially affects livestock production, wildlife conservation, interactive effects on tree and grass dynamics. A variety predator–prey interactions, nutrient cycling, and carbon storage (Scholes and Archer 1997, Angassa and Baars 2000, van Auken 2000, Bond and Keeley 2005, Riginos Manuscript received 17 June 2013; revised 16 August 2013; and Grace 2008). In particular, there is a lack of accepted 11 September 2013. Corresponding Editor: N. T. Hobbs. controlled replicated experimental studies that quantify 7 Corresponding author. E-mail: [email protected] (1) the separate and combined effects of domestic and 741 742 DUNCAN M. KIMUYU ET AL. Ecological Applications Vol. 24, No. 4 native herbivores on fuel loads and fire temperatures loads, (2) what are the patterns of ant colony survival and (2) the separate and combined effects of fire and after burning, and (3) do differences in ant colony subsequent herbivory on the survival of savanna trees. survival and subsequent shifts in ant species composition We address the former and describe the experimental or occupancy correlate with herbaceous fuel loads and design that will be used to test the latter. burn temperatures? A number of previous results have provided partial evidence for these interactions. For example, there is STUDY SITE AND METHODS some evidence from African savannas (Hobbs 1996, Study site and species O’Connor et al. 2011), North American grasslands (van This research was carried out in Acacia drepanolobium Auken 2000), and Australian savannas (Leonard et al. wooded savanna on the Laikipia Plateau, Kenya (368520 2010) that preburn grazing decreases fire frequencies and E, 08170 N; 1810 m above sea level). This plant intensities. This suggests feedback loops between her- community overlies high clay black cotton soils and is bivory and fire that current models of savanna dynamics representative of similar ecosystems that occur exten- do not include (Sankaran et al. 2004). However, most sively throughout eastern and southern Africa. Rainfall existing information on the relationship among herbi- averages 550–600 mm at our study site and is weakly vores, herbaceous fuel load, and fire behavior is based trimodal, with a distinct dry season from December to on anecdotal observations (Nader et al. 2007), studies March. involving only domesticated herbivores (Savadogo et al. The whistling thorn tree, Acacia drepanolobium, 2007, Gambiza et al. 2008, Davies et al. 2010), studies accounts for 97% of woody cover at our study site that do not control or monitor preburn grazing (Kerby (Young et al. 1998). At some branch nodes, A. et al. 2007), or those lacking replication (Leonard et al. drepanolobium produces hollow swollen thorns that 2010). serve as ant domatia. The trees also produce extrafloral Ant mutualisms may mediate fire–herbivore interac- nectaries at the bases of leaves to nourish symbiotic ants tions. Symbiotic ant species that reside on Acacia (Hocking 1970, Huntzinger et al. 2004). At our study drepanolobium trees are effective in reducing browsing site, each tree is occupied by one of four species of ants: by megaherbivores (Madden and Young 1992, Palmer et Crematogaster mimosae, C. sjostedti, C. nigriceps,or al. 2008a, Martins 2010, Stanton and Palmer 2011), Tetraponera penzigi. These ant species differ in various increasing individual tree fitness (Palmer and Brody aspects of their behavior, including the average size and 2007, Goheen and Palmer 2010, Palmer and Brody number of trees they occupy, their relative competitive 2013), and stabilizing tree cover across the landscape abilities, their benefits to host trees (Young et al. 1997, (Goheen and Palmer 2010). To the extent that fire may Stanton and Palmer 2011), and their behavioral weaken or destroy ant colonies (Palmer et al. 2008b), it responses to fire. During a fire, C. nigiceps and C. may lead to subsequent shifts in ant species composition mimosae evacuate domatia and take refuge in insulated or declines in ant abundance across the landscape, either cracks in the soil (Jaffe and Isbell 2009; T. Palmer, of which could make trees in burned areas more personal communication), reoccupying the tree after the susceptible to browsing and the destructive effects of fire. Tetraponera penzigi do not leave their domatia, and elephants, with subsequent cascades to other herbivore C. sjostedi take refuge in holes on tree stems, often and plant guilds. created by Cerambycidae larvae (Palmer et al. 2008a). In order to tease apart these complex dynamics The Kenya Long-term Exclosure Experiment (KLEE) among diverse herbivores (wild and domestic, grazing is located at the Mpala Research Centre. Since 1995, we and browsing), fire, savanna grasses, trees, and acacia have been manipulating the presence and absence of ants, we implemented prescribed burns inside the three guilds of large herbivores: livestock (cattle, C), replicated Kenya Long-term Exclosure Experiment wildlife (large mammals 15–1000 kg, W), and mega- (KLEE) in early 2013. By experimentally burning within herbivores (elephants and giraffes, M). There are 18 different herbivore treatments, we were able to simulta- plots, each 200 3 200 m, representing three replicate neously test the effects of a diverse guild of herbivores blocks of each of six combinations of large herbivores on woody and herbaceous fuel loads and fire tempera- (Fig. 1). For details of the experimental design, see tures and will be able to quantify how postburn foraging Young et al. (1998). by herbivores may be responsible for the delayed mortality of larger trees after a fire. To our knowledge, Controlled burns this is the first fully crossed, replicated, field experiment In February and March 2013, we burned one 30 3 30 that independently manipulates fire and multiple guilds m subplot in each of the 18 4-ha KLEE plots (Fig. 1). of domestic and wild large herbivores (but see Collins Each burn therefore covered 2.25% of each plot. These and Smith 2006, Collins and Calabrese 2012, Koerner subplots were situated within each plot using the and Collins 2014). following criteria: (1) subplots were located along the We address the following questions: (1) how do plot boundary that was most accessible for fire-fighting different guilds of herbivores indirectly affect burn equipment, (2) local sources of landscape heterogeneity temperatures by influencing woody and herbaceous fuel (e.g., termite mounds and old livestock enclosure sites) June 2014 HERBIVORES, FIRE, TREES, AND ACACIA ANTS 743 were avoided, and (3) we attempted to select subplots that were similar with respect to the density and size structure of Acacia drepanolobium trees (.1 m tall). Understory composition was similar across these sites (see Young et al. 1998) with the exception of the plots excluding all large herbivores, which have experienced increases in forb cover and shifts in the relative abundances of the five dominant grass species due to KLEE treatment effects. We refrained from grazing cattle in the KLEE plots for five months (one rainy and one dry period) prior to the burns. In early 2013, we estimated herbaceous fuel loads by harvesting all aboveground herbaceous biomass in three randomly located 1 3 1 m quadrats adjacent to each designated burn plot. This material was air dried to constant mass and weighed. Every tree (all size classes) in each burn plot was permanently tagged and its height and stem diameter measured at 15 cm (for trees .1 m tall) or at the base (for trees ,1 m tall). A total of 4304 trees were tagged and measured (mean per plot, 240; range, 130–359). In addition, the species identity of each tree’s resident ant was recorded, as was evidence of the presence of stem-boring cerambycid beetle damage. In 2011, we documented the effects that elephants had on Acacia drepanolobium densities by carrying out surveys in each of the 18 KLEE plots, six of which allowed elephants and 12 of which did not (see Fig. 1). We counted all trees within three 100 3 4 m transects (1200 m2 total per plot). Trees were categorized into height classes: ,1.00 m, 1.00–1.99 m, 2.00–2.99 m, FIG. 1. Schematic of the Kenya Long-term Exclosure Experiment (KLEE) plots and the burned subplots (solid 3.00–3.99 m, and 4.00 m). We did not survey the burn squares) within each. The letters inside each plot indicate the subplots for fine woody debris before the burns, so large herbivores allowed: C, cattle; M, megaherbivores (ele- instead we documented the general relationship between phants and giraffes); W, nonmegaherbivore wildlife .15 kg; O, tree density and fine woody debris. In July 2013, along all large herbivores excluded. 10 of the same (unburned) transects surveyed for tree density in 2011, we collected and weighed all fine woody subplots, either an access road or cleared fence lines debris (FWD; dead woody material on the ground with provided firebreaks. Over a period of three days (28 diameters ,2.0 cm) within 1 m of each transect line (a February–2 March 2013), a single burn was completed 100 3 2 m transect). in each of the 18 subplots. Burn boundaries were wet- We monitored fire temperatures with ceramic tiles lined and then back-burned to create 5–10 m of firebreak painted with Tempilac (LA-CO Industries, Elk Grove along downwind edges. The interiors of all plots were Village, Illinois, USA) paints. We used paints designed burned using head fires with two technicians simulta- to melt at each of the following six temperatures: 808C neously lighting the upwind edges to create a perimeter (1758F), 1508C (3008F), 2608C (5008F), 4008C (7508F), ring fire. Flame heights ranged from less than 1 m to a 5108C (9508F), and 6208C (11508F). We placed 18 tiles in maximum of 3–4 m. From first ignition to the end of the each burn plot: two in each of three grassy areas away burn, each lasted between 8 and 13 minutes (two-thirds from trees and four on each of three separate, randomly of the burns were 9–11 minutes). All burns consumed chosen trees. In each open grassy sampling location, we virtually all of the understory vegetation (save some placed one tile on the ground (under any leaf litter tussock bases) and fine woody debris. Each day’s present) and attached one to a piece of rebar at 50–60 burning began at ;08:00 (07:47–08:06) and finished cm above the ground. At each tree, we placed one tile on approximately midday (11:17–12:15). the ground (under any leaf litter present) and attached On each burn day we periodically recorded air three tiles to the tree, one at 100 cm, one at 180–200 cm, temperature, wind temperature, and relative humidity and one at 270–300 cm above the ground. (Appendix B). Air temperatures during the burns ranged Firebreaks were established between 13 January and 6 from 158 to 288C, increasing during the course of each February 2013. A 1–2 m wide swath was slashed to a day. Winds ranged from 3.5 to 12.5 km/h, increasing vegetation height of 5–10 cm. In addition, for many during the course of each day. Relative humidity ranged 744 DUNCAN M. KIMUYU ET AL. Ecological Applications Vol. 24, No. 4

TABLE 1. ANOVA for the effects of cattle, wildlife, and block on the densities of trees .1 m tall, mean herbaceous fuel loads, and mean minimum ﬁre temperatures across all tile locations (heights).

Tree density Herbaceous fuel load Minimum temp Source df FP F P F P Block 2 9.32 0.0004 7.41 0.008 3.74 0.055 Cattle 1 0.33 0.57 28.7 0.0002 6.91 0.022 Wildlife 2 6.14 0.004 9.95 0.003 3.65 0.058 Error 48 Notes: Tree density data are from the overall plots, not the burn subplots, which were selected to reduce variation in tree density. Mean fuel loads and minimum fire temperatures are from burn subplots. from 80% to 30%, declining during the course of each for tree size classes that were relatively consistent for day. mean survivorship and had sufficient sample sizes to To quantify the presence or absence of acacia ants in estimate proportions reliably. Ant colony survivorships the first week after the burns, we carried out a survey of were compared with chi-square contingency tests. trees in the burned plots for a random subset of 30–50 tagged trees in each plot. Data on the fate of individual RESULTS trees in the burns could not be collected at that time Before the burns, experimental exclusion of mega- because rains had not yet fallen that would result in herbivores (elephants and giraffes) had resulted in 42% either a flush of leaves from aboveground shoots or the more trees .1 m tall (1257 trees/ha) than in plots where coppicing of top-killed trees. (These data will be all large wildlife was allowed access (884 trees/ha). In reported in subsequent publications). addition, there were significant block effects, with trees At the same time, we collected the ceramic tiles and becoming less dense in blocks farther north (Table 1). scored them for the lowest temperature class of paint Our attempts to reduce these sources of variation in tree melted by the fire (minimum fire temperature). If none of density in selecting our burn subplots were only partly the paints melted, a default of 208C (air temperature) successful. Although the number of trees .1 m tall did was used for analysis. For 12 tiles (,4% of the total), the not significantly differ across our wildlife treatments in higher temperature paints were charred and not our selected burn subplots (P ¼ 0.34), there remained readable. These may have been the hottest tiles; all but significant differences across blocks (P ¼ 0.002). one were at ground level, and nine out of these 11 were At ground level, minimum temperatures were on under trees. This represented 17% of the tiles at ground average 188C higher under trees than in open grassy level under trees, and our minimum temperature areas. On trees, fire temperatures declined with greater estimates for these locations should be considered height from the ground (Fig. 2a). Even at 3 m, however, conservative. more than half the tiles reached temperatures of at least 808C. In grassy areas between trees, fire temperatures Statistical analyses were nearly as high at 0.5 m as at ground level (Fig. 2a). Tree densities, mean herbaceous fuel loads, mean fire Differences in mean minimum fire temperatures were temperatures, and ant survivorship (percentage of trees associated with wildlife treatment, cattle treatments, and retaining their ants) were calculated for each of the 18 block effects (Table 1). Analysis of the data on preburn burns and were analyzed with ANOVA (including block tree density revealed that the block effects in fire as a fixed effect; N, C, and S for North, Central, and temperatures were strongly associated with the densities South blocks), testing the effects of herbivore treatments of trees .1 m tall (but not smaller trees). The more trees in a 2 3 3 complete factorial design (two cattle (.1 m), the hotter the fire (r2 ¼ 0.63, P , 0.001; Fig. 2b). treatments [C, no cattle] crossed with three herbivore Herbaceous fuel loads were not significantly correlated treatments [MW, W, no wildlife]). Interaction terms with the densities of trees (r2 ¼ 0.04, P ¼ 0.45). Fine were not significant, so were not included in the model. woody debris was found almost entirely directly beneath Correlations were calculated among tree density, larger trees and was strongly correlated at the transect mean fire temperature, and fine woody debris. Correla- level with tree density in unburned plots (r2 ¼ 0.83, P ¼ tions between herbaceous fuel loads and minimum fire 0.0002, Fig. 2c). temperatures were calculated separately for each of the After controlling for block effects (strongly correlated three blocks (which differed significantly in tree densities with tree density) mean minimum fire temperatures were and mean herbaceous fuel loads; Table 1) and the also positively correlated with mean herbaceous fuel overall pattern tested with a Fisher’s combined proba- loads across plots (P , 0.01). Both mean herbaceous bility test. fuel loads and mean minimum fire temperatures were Ant colony survivorships (proportion of trees retain- significantly affected by herbivore treatments (Table 1). ing their ant colonies) within ant species were calculated Herbaceous fuel loads were lower and fires burned June 2014 HERBIVORES, FIRE, TREES, AND ACACIA ANTS 745

cooler in plots to which cattle had access (C , O, WC , W, and MWC , MW) and in plots to which more wildlife guilds had access (MW , W , O and MWC , WC , C, Fig. 3). Across all ant species, survivorship on trees was negatively correlated with mean herbaceous fuel load and mean minimum fire temperature across herbivore treatments for trees 1–2 m tall (herbaceous fuel load, r2 ¼ 0.71, P ¼ 0.04; temperature, r2 ¼ 0.73, P ¼ 0.03), but not for trees 2–4 m tall (both P . 0.50). Before the burns, virtually all trees .50 cm tall were inhabited by ants. After burning, trees ,50 cm tall were often burned completely aboveground, and trees ,1m tall were almost all uninhabited by ants a week after the fire, many of them having all their swollen thorns burned off (D. M. Kimuyu and T. P. Young, personal observations). For trees .1 m, the proportion of trees occupied by ants in the first week after the burns increased with tree height, but even trees .5 m tall sometimes had no detectable ant presence (Fig. 4). Ant survival also differed significantly by ant species, even after controlling for tree size (Table 2). Tetraponera colonies rarely survived fire (12%). Crematogaster mimosa and C. nigiceps had relatively high survivorships (68–91%), and C. sjostedti had intermediate survivorship (37–68%). Among trees originally occupied by C. sjostedti, those that had evidence of cerambycid beetle stem boring were significantly more likely to retain their ant colonies than those that did not have beetle damage (chi-square ¼ 33.66, P , 0.001). This benefit was most pronounced in smaller trees (Fig. 5). Lepisota canescens, a ground-

FIG. 2. (a) Mean minimum fire temperature as a function of height aboveground for tiles placed at two heights in grassy areas away from trees and for tiles placed beneath and at different height in trees. (b) Mean minimum fire temperatures as a function of the preburn density of trees .1 m tall. (c) The abundance of fine woody debris (,2 cm diameter) as a function FIG. 3. Mean minimum fire temperatures as a function of of the density of tree .1 m tall along transects in unburned herbivore treatment. Plots with cattle burned cooler than plots areas. without cattle, and plots with more native herbivore guilds burned cooler than plots with fewer (MW, MWC , W, WC , O, C). See Fig. 1 caption for herbivore treatments. Bars represent one standard error. 746 DUNCAN M. KIMUYU ET AL. Ecological Applications Vol. 24, No. 4

effects, resulting in different fire temperatures. These data strongly suggest that the presence of large herbivores (particularly elephants), reduces fire temperatures by reducing the densities of trees. Why did subplots with greater densities of tall trees (but not small trees) burn hotter? We suggest two possibilities. First, grass biomass tends to be higher under trees. Although our subplot surveys failed to reveal a positive relationship between tree density and herbaceous fuel load, the relatively small sample size of herbaceous biomass (three quadrats per subplot) may not have picked up on the tree biomass under trees. Perhaps more likely, the hotter fires in subplots with more trees may have been due to the fine woody debris that collects under and around these trees. We did not survey fine woody debris in the burn subplots prior to burning, but our survey across the broader KLEE plots did show a strong positive relationship between tree density and fine woody debris (Fig. 2c). Such fine woody debris was essentially lacking after the burns, presum- FIG. 4. Ant survivorship as a function of A. drepanolobium tree size across all ant species. This does not include trees ably consumed by fire. completely consumed by the fire, of which there were many in In addition, both herbaceous fuel loads and fire the 0–1 m height class. temperatures were reduced in plots grazed by livestock and/or wild herbivores (Fig. 3). The fact that we found dwelling ant species that is not otherwise observed on A. significant effects of previous cattle grazing even after drepanolobium, was found on 13 out of 595 resurveyed resting the plots for one growing (wet) season demon- trees after the burns and in one case was observed strates that the influence of grazing on herbaceous fuel removing ant brood from a burned tree. loads and fire behavior extends beyond the immediate effect of grazers removing current season’s plant growth. DISCUSSION Consistent with the observation that grass biomass is Fire temperatures and herbivory higher under A. drepanolobium trees (Riginos et al. 2009) and perhaps also because of fine woody debris, we found It appears that both browsing and grazing herbivores, greater ground-level fire temperatures under trees than both wild and domestic, can have significant effects on away from trees. This is also likely to put trees at greater fire temperatures in this A. drepanolobium wooded risk of fire damage. savanna, at least for the controlled burns we carried out in February and March 2013. Minimum fire Fire and symbiotic acacia ants temperatures were strongly and positively associated One of our key findings is the indirect effect that with densities of A. drepanolobium trees .1 m tall (Fig. herbivores have on ant mortality due to fire, mediated 2b). In the plots overall, megaherbivores have dramat- by browser and grazer effects on tree densities and ically reduced densities of these trees. Although we herbaceous fuel loads. The vulnerability of symbiotic successfully controlled for this variance across wildlife ants to fire varied greatly across the four ant species in treatments in our selection of burn subplots, there was our study (Appendix A). The high colony survival of C. still considerable variation, mostly associated with block nigiceps and C. mimosae may be attributable to colonies

TABLE 2. Ant survivorship on trees of different heights (percentage of trees of a given height occupied by a given ants species that retained their ant colony).

Tree height N Tetraponera Crematogaster nigiceps C. mimosae C. sjostedti Chi-square P 1–2 m 132 5% (1/24) 91% (20/22) 70% (37/53) 37% (10/27) 44.52 ,0.0001 2–4 m 372 13% (7/52) 82% (23/28) 75% (122/163) 52% (65/124) 69.98 ,0.0001 4 m 91 0% (0/5) no data 68% (26/38) 68% (30/47) 8.91 0.01 Chi-square 2.18 0.23 0.22 1.05 P 0.34 0.63 0.64 0.31 Note: N is the number of trees. The ratios in parentheses are the number of trees occupied by ants that retained their ant colony divided by the number of trees that had an ant colony to begin with. The chi-square tests to the right compare ant colony survivorships within tree height classes across ant species. The chi-square tests on the bottom compare ant colony survivorships across tree height classes within ant species. The estimate for C. mimosae on taller trees includes four trees whose surviving ants in treetops could not be reliably identified to species. June 2014 HERBIVORES, FIRE, TREES, AND ACACIA ANTS 747 of these species rapidly evacuating domatia downwind of the fire front and taking refuge in insulated cracks in the soil (T. Palmer, personal communication), presum- ably returning after the fire. Although C. sjostedti take refuge in holes created by Cerambycidae larvae on tree stems, these holes are rare among trees with ,4.0 cm stem diameter (usually ,2 m tall trees; Palmer et al. 2013), which are most vulnerable to ground fire. In this study, most C. sjostedti colonies occupying short trees without these holes succumbed to fire (Fig. 5). Tetraponera penzigi, which had the lowest colony survival, usually inhabits shorter trees and does not evacuate colonies in the event of fire, both of which make them more vulnerable to the high ground fire temperatures. The two ant species that exhibited the greatest survival after fire (C. nigiceps and C. mimosae) are also the two species that are most effective in aggressively defending trees against herbivory (Palmer and Brody 2007). Our prescribed burns caused significant ant colony mortality, especially on smaller trees. It is possible that FIG. 5. Proportion of trees originally occupied by Crema- there will be further reductions in ant occupancy in the togaster sjostedti that retained their ant colonies after the burns based on whether or not the tree had preburn evidence of future if trees in burns are physiologically stressed and damage by stem-boring cerambycid beetles. Beetle damage was produce fewer of the extrafloral nectaries upon which rare in trees ,250 cm tall. most resident ant species depend for nourishment (Hocking 1970). Conversely, we expect trees that have 2004, 2010, Maclean et al. 2011). This research suggests lost their ant colonies to eventually be colonized by two additional indirect pathways: wild and domestic founder queens or by colony expansion from nearby large herbivores affect burn temperatures that could trees. Long-term monitoring will reveal such coloniza- affect postfire tree demography (to be revealed in later tion and whether there are spatially explicit or species- surveys), and fire also affects the ant symbionts on which specific patterns to such recolonization. these trees depend. For A. drepanolobium, the loss or reduction of ant Fireandgrazingcanbeemployedtoachieve colonies could have far-reaching implications for the complementary range management goals (Kirkpatrick protection of the trees against herbivory. In previous et al. 2011). For example, in the western United States, burns near the study site, most of the tall trees (.2m) Diamond et al. (2009) demonstrated that targeted were able to survive the immediate effects of fire, but grazing and prescribed burning can reduce the biomass there was an unexpected increase in the number of large and cover of cheatgrass (Bromus tectorum), resulting in trees toppled by elephants and killed (R. L. Sensenig, reductions in flame length and rate of spread in a system unpublished data) and a striking decline in the density of where fire is a recent (and damaging) anthropogenic larger trees (Okello et al. 2008) in burn sites relative to unburned areas in the years following burning. This disturbance. Integrated burning and grazing has also suggests several possibilities: (1) fire has a delayed direct been used to manipulate habitat heterogeneity (Fuhlen- effect on tree mortality, (2) fire makes trees more dorf and Engle 2004), to restore degraded habitats, and susceptible to elephant damage, or (3) elephants respond increase biodiversity (Fuhlendorf and Engle 2001). to fire in a way that increases their impact on trees. We These data also have implications for the conservation know that elephants preferentially select A. drepano- of large mammals in savanna ecosystems. Although the lobium trees from which ants have been experimentally impacts of herbivory are analogous to the impacts of fire removed (Goheen and Palmer 2010). Do elephant in their removal of herbaceous biomass (Bond and similarly attack trees that have lost their ant colonies Keeley 2005), they have very different effects on long- through fire? Continued monitoring of our experimental term tree and grass cover and community ecological design can test the interaction between fire and elephants interactions. Our data suggest that an additional cost of and whether ants mediate this interaction. the loss of meso- and megafaunal wildlife is that it increases savanna fuel loads and fire temperatures (see Implications for management and conservation also Gill et al. 2009). This results in increased symbiotic We have previously demonstrated in this system that ant mortality and likely other expressions of increased livestock and different guilds of wildlife have complex fire severity. There are parallels to the dynamics of some direct and indirect effects on the recruitment and of the coniferous forests in the western United States, survival of Acacia drepanolobium trees (Goheen et al. where human intervention (there, in the form of fire 748 DUNCAN M. KIMUYU ET AL. Ecological Applications Vol. 24, No. 4 suppression) has also increased the likelihood of high- fuel loads in Baikiaea plurijuga Harms woodland in severity fires (Moore et al. 1999, Savage and Mast 2005). northwestern Zimbabwe. African Journal of Ecology In our system, the replacement of native grazers by 46:637–645. Gill, J. L., J. W. Williams, S. T. Jackson, K. B. Lininger, and cattle only partly compensates for the loss of large G. S. Robinson. 2009. Pleistocene megafaunal collapse, novel mammal biodiversity, reducing the herbaceous fuels, but plant communities, and enhanced fire regimes in North not the woody fuels reduced by wildlife. America. Science 326:1100–1103. The results presented here are the first stage in a long- Gillson, L. 2004. Evidence of hierarchical patch dynamics in an east African savanna? Landscape Ecology 19:883–894. term monitoring program that will further elucidate the Goheen, J. R., F. Keesing, B. F. Allan, D. Ogada, and R. S. effects of fire and large mammal herbivory on each other Ostfeld. 2004. Net effects of large mammals on Acacia and on the savanna ecosystem in which they occur. seedling survival in an African savanna. Ecology 85:1555– Already, we have documented both multitrophic inter- 1561. actions and herbivore–fire feedbacks that are likely to Goheen, J. R., and T. M. Palmer. 2010. Defensive plant-ants stabilize megaherbivore-driven landscape change in an play out in fascinating ways. African savanna. Current Biology 20:1768–1772. Goheen, J. R., T. M. Palmer, F. Keesing, C. Riginos, and T. P. ACKNOWLEDGMENTS Young. 2010. Large herbivores facilitate savanna tree We thank the fire crew team, especially Fredrick Erii, John establishment via diverse and indirect pathways. Journal of Lochukuya, Mathew Namoni, Jackson Ekadeli, Wilson Lon- Animal Ecology 79:372–382. gor, Naisikie John, James Ekiru, Simon Lima, and Boniface Hobbs, N. T. 1996. Modification of ecosystems by ungulates. Kirwa. Essential logistical support was provided by Mpala Journal of Wildlife Management 60:695–713. Farm (Michael Littlewood) and Mpala Research Centre Hocking, B. 1970. Insect associations with swollen thorn (Margaret Kinnaird, Alick Roberts, and Julius Nakalonyo). acacias. Transactions of the Royal Entomological Society Useful comments came from Sally Koerner and two anony- of London 122:211–255. mous reviewers. Funding support came from the National Holdo, R. M., R. D. Holt, and J. M. Fryxell. 2009. Grazers, Science Foundation (LTREB BSR 08-16453 and 13-56034), browsers, and fire influence the extent and spatial pattern of National Geographic Society (Grant 9106-12), and Goshen tree cover in the Serengeti. Ecological Applications 19:95– College to T. P. Young, R. L. Sensenig, K. E. Veblen, C. 109. Riginos, and K. Caylor. Huntzinger, P. M., R. Karban, T. P. Young, and T. M. Palmer. 2004. Relaxation of induced indirect defenses of acacias LITERATURE CITED following removal of mammalian herbivores. Ecology Allred, B. W., S. D. Fuhlendorf, D. M. Engle, and R. D. 85:609–614. Elmore. 2011. Ungulate preference for burned patches Jaffe, K. E., and L. A. Isbell. 2009. After the fire: benefits of reveals strength of fire–grazing interaction. Ecology and reduced ground cover for vervet monkeys (Cercopithecus Evolution 1:132–144. aethiops). American Journal of Primatology 71:252–260. Angassa, A., and R. M. T. Baars. 2000. Ecological condition of Kerby, J. D., S. D. Fuhlendorf, and D. M. Engle. 2007. Land- encroached and non-encroached rangelands in Borana, scape heterogeneity and fire behavior: scale-dependent feed- Ethiopia. African Journal of Ecology 38:321–328. back between fire and grazing processes. Landscape Ecology Archibald, S., W. J. Bond, W. D. Stock, and D. H. K. 22:507–516. Fairbanks. 2005. Shaping the landscape: fire–grazer interac- Kirkpatrick, J. B., J. B. Marsden-Smedley, and S. W. J. Leonard. tions in an African savanna. Ecological Applications 15:96– 2011. Influence of grazing and vegetation type on post-fire 109. flammability. Journal of Applied Ecology 48:642–649. Baxter, P. W. J., and W. M. Getz. 2005. A model-framed Koerner, S. E., and S. L. Collins. 2014. Interactive effects of evaluation of elephant effects on tree and fire dynamics in grazing, drought, and fire on grassland plant communities in African savannas. Ecological Applications 15:1331–1341. North America and South Africa. Ecology 95:98–109. Bond, W. J., and J. E. Keeley. 2005. Fire as a global ‘herbivore’: Leonard, S. W. J., J. B. Kirkpatrick, and J. B. Marsden- the ecology and evolution of flammable ecosystems. Trends Smedley. 2010. Variation in the effects of vertebrate grazing in Ecology and Evolution 20:387–394. on fire potential between grassland structural types. Journal Collins, S. L., and L. B. Calabrese. 2012. Effects of fire, grazing of Applied Ecology 47:876–883. and topographic variation on vegetation structure in tallgrass Levick, S. R., G. P. Asner, T. Kennedy-Bowdoin, and D. E. prairie. Journal of Vegetation Science 23:563–575. Knapp. 2009. The relative influence of fire and herbivory on Collins, S. L., and M. D. Smith. 2006. Scale-dependent savanna three-dimensional vegetation structure. Biological interaction of fire and grazing on community heterogeneity Conservation 142:1693–1700. in tallgrass prairie. Ecology 87:2058–2067. Maclean, J. E., J. R. Goheen, D. F. Doak, T. M. Palmer, and Diamond, J. M., C. A. Call, and N. Devoe. 2009. Effects of T. P. Young. 2011. Cryptic herbivores mediate the strength targeted cattle grazing on fire behavior of cheatgrass- and form of ungulate impacts on a long-lived savanna tree. dominated rangeland in the northern Great Basin, USA. Ecology 92:1626–1636. International Journal of Wildland Fire 18:944–950. Madden, D., and T. P. Young. 1992. Symbiotic ants as an Dublin, H. T., A. R. E. Sinclair, and J. McGlade. 1990. alternative defense against giraffe herbivory in spinescent Elephants and fire as causes of multiple stable states in the Acacia drepanolobium. Oecologia 91:235–238. Serengeti-Mara woodlands. Journal of Animal Ecology Mapiye, C., M. Mwale, N. Chikumba, and M. Chimonyo. 59:1147–1164. 2008. Fire as a rangeland management tool in the savannas Fuhlendorf, S. D., and D. M. Engle. 2001. Restoring of southern Africa: a review. Tropical and Subtropical heterogeneity on rangelands: ecosystem management based Agroecosystems 8:115–124. on evolutionary grazing patterns. BioScience 51:625–632. Martins, D. J. 2010. Not all ants are equal: obligate acacia ants Fuhlendorf, S. D., and D. M. Engle. 2004. Application of the provide different levels of protection against mega-herbi- fire–grazing interaction to restore a shifting mosaic on vores. African Journal of Ecology 48:1115–1122. tallgrass prairie. Journal of Applied Ecology 41:604–614. McNaughton, S. J., R. W. Ruess, and S. W. Seagle. 1988. Large Gambiza, J., B. M. Campbell, S. R. Moe, and I. Mapaure. mammals and process dynamics in African ecosystems. 2008. Season of grazing and stocking rate interactively affect BioScience 38:794–800. June 2014 HERBIVORES, FIRE, TREES, AND ACACIA ANTS 749

Moncrieff, G. R., L. M. Kruger, J. J. Midgley, and S. J. nation of assumptions and mechanisms invoked in existing McNaughton. 2008. Stem mortality of Acacia nigrescens models. Ecology Letters 7:480–490. induced by the synergistic effects of elephants and fire in Sankaran, M., J. Ratnam, and N. Hanan. 2008. Woody cover Kruger National Park, South Africa. Journal of Tropical in African savannas: the role of resources, fire and herbivory. Ecology 24:655–662. Global Ecology and Biogeography 17:236–245. Moore, M. M., W. W. Covington, and P. Z. Fule. 1999. Savadogo, P., D. Zida, L. Sawadogo, D. Tiveau, M. Tigabu, Reference conditions and ecological restoration: a south- and P. C. Oden. 2007. Fuel and fire characteristics in western ponderosa pine perspective. Ecological Applications savanna-woodland of West Africa in relation to grazing and 9:1266–1277. dominant grass type. International Journal of Wildland Fire O’Connor, T. G., C. M. Mulqueeny, P. S. Goodman. 2011. 16:531–539. Determinants of spatial variation in fire return period in a Savage, M., and J. N. Mast. 2005. How resilient are semiarid African savanna. International Journal of Wildland southwestern ponderosa pine forests after crown fires? Fire 20:540–549. Okello, B. D., T. P. Young, C. Riginos, D. Kelly, and T. Canadian Journal of Forest Research 35:967–977. O’Connor. 2008. Short-term survival and long-term mortal- Scholes, R. J., and S. R. Archer. 1997. Tree-grass interactions in ity of Acacia drepanolobium after a controlled burn in savannas. Annual Review of Ecology and Systematics Laikipia, Kenya. African Journal of Ecology 46:395–401. 28:517–554. Palmer, T. M., and A. K. Brody. 2007. Mutualism as reciprocal Sensenig, R. L., M. Demment, and E. A. Laca. 2010. exploitation: ant guards defend foliar but not reproductive Allometric scaling predicts preferences for burned patches structures of an African ant-plant. Ecology 88:3004–3011. in a guild of East African grazers. Ecology 91:2898–2907. Palmer, T. M., and A. K. Brody. 2013. Enough is enough: the Stanton, M. L., and T. M. Palmer. 2011. The high cost of effects of symbiotic ant abundance on herbivory, growth, and mutualism: effects of four species of East African ant reproduction in an African acacia. Ecology 94:683–691. symbionts on their myrmecophyte host tree. Ecology Palmer, T. M., M. L. Stanton, T. P. Young, J. R. Goheen, 92:1073–1082. R. M. Pringle, and R. Karban. 2008a. Breakdown of an ant- Trollope, W. S. W. 2011. Personal perspectives on commercial plant mutualism following the loss of large herbivores from versus communal African fire paradigms when using fire to an African savanna. Science 319:192–195. manage rangelands for domestic livestock and wildlife in Palmer, T. M., M. L. Stanton, T. P. Young, J. R. Goheen, southern and East African ecosystems. Fire Ecology 7:57–72. R. M. Pringle, and R. Karban. 2008b. Putting ant-acacia van Auken, O. W. 2000. Shrub invasions of North American mutualisms to the fire: response. Science 319:1760–1761. semiarid grasslands. Annual Review of Ecology and System- Palmer, T. M., M. L. Stanton, T. P. Young, J. S. Lemboi, J. R. atics 31:197–215. Goheen, and R. M. Pringle. 2013. A role for indirect van Langevelde, F., et al. 2003. Effects of fire and herbivory on facilitation in maintaining diversity in a guild of African the stability of savanna ecosystems. Ecology 84:337–350. acacia ants. Ecology 94:1531–1539. Riginos, C., and J. B. Grace. 2008. Tree density affects wild Waldram, M. S., W. J. Bond, and W. D. Stock. 2008. ungulate habitat use and herbaceous community character- Ecological engineering by a mega-grazer: white rhino impacts istics in a Kenyan savanna. Ecology 89:2228–2238. on a South African savanna. Ecosystems 11:101–112. Riginos, C., J. B. Grace, D. J. Augustine, and T. P. Young. Young, T. P., B. Okello, D. Kinyua, and T. M. Palmer. 1998. 2009. Local versus landscape-scale effects of savanna trees on KLEE: a long-term multispecies herbivore exclusion exper- grasses. Journal of Ecology 97:1337–1345. iment in Laikipia, Kenya. African Journal of Range and Sankaran, M., et al. 2005. Determinants of woody cover in Forage Science 14:94–102. African savannas. Nature 438:846–849. Young, T. P., C. H. Stubblefield, and L. A. Isbell. 1997. Ants Sankaran, M., J. Ratnam, and N. Hanan. 2004. Tree–grass on swollen-thorn acacias: species coexistence in a simple coexistence in savannas revisited—insights from an exami- system. Oecologia 109:98–107.

SUPPLEMENTAL MATERIAL

Appendix A Ant survivorship (across all ant species) in different herbivore treatments (Ecological Archives A024-044-A1).

Appendix B Air temperature, relative humidity, and wind speed with respect to time of day (Ecological Archives A024-044-A2).